Trend testing of grey dynamic models

The testing method of the fitting accuracy of grey dynamic models is mainly studied in this paper. First the existing testing methods, e.g., residuals testing and posteriori deviation testing, are briefly introduced and evaluated, and the merits and defects of the existing testing methods are pointed out, e.g., the existing testing method indeed do not consider the quality for grey models to fit for the trend of data sequences. Then the features, which a kind of testing method for the fitting accuracy of grey model should possess, are analyzed. At last, based on the specific features of grey models, a new kind of testing method for the fitting accuracy of grey dynamic model is designed, that is, trend testing, which may reflect the following aspects: (1) the extent of fitting the whole trend of data sequences; (2) the extent of fitting each point of data sequences; (3) the comparability of results among different problems.